提示词模板
Photo → Chinese Shadow-Puppet Split Poster (GPT Image, Chinese Prompt)
A Chinese GPT Image prompt that turns each photo you upload into a 3:4 poster: the top half keeps your original photo, the bottom half re-carves its subject as a traditional Chinese shadow puppet (皮影) glowing on a warm, back-lit paper screen with lots of empty space. Shared by @daaihq.
从示例开始
点一张示例图,整套设置立即套用,再改成你自己的。
改成你的
只展示值得改的部分,提示词会实时更新。
你的提示词
请将我上传的每一张照片分别制作成一张独立的高级设计海报,不多图拼接,每张照片单独输出。整体采用 3:4 竖版构图,上下两个区域高度严格 1:1,各占画面 50%。 上半部分保留原始照片,保持主体身份、结构、姿态、真实质感、自然光影与原有色彩氛围,仅进行轻微高级调色,使其具有艺术摄影与东方展览图像质感。为适配画幅,可自然扩展环境背景,但不得拉伸、扭曲或改变主体。 下半部分提取照片中最具识别性的**主体、轮廓、姿态与叙事关系**,重构为一具**中国传统皮影戏偶 / shadow puppet**,在一面被暖光照亮的半透明宣纸幕布上演出。无论原图是人物、动物、建筑、植物、器物还是风景,都被理解为一个“可以被操纵的角色”:主体被拆分为少量关节部件,以铆钉相连,姿态取自原照片中最有表情的那个瞬间,并配有一两根细长的操纵竹签自画面下方伸入。 不要写实描绘,不要卡通化,而是用皮影的造型语法重新概括:侧面化、平面化、装饰性镂空、云纹与几何雕刻纹样、局部的半透明彩染。保留原物最具识别度的特征,使观者一眼认出,同时感到它被一位老手艺人重新理解和雕刻过。 光影是这件作品的灵魂:幕布后方一盏暖色灯光,使皮影的镂空处透出温暖光点,彩染部分呈现半透明的朱红、靛青、橄榄绿与琥珀黄,皮影边缘带有柔和的投影虚边和轻微光晕,仿佛正在缓慢移动。可以有极轻微的双影或动态虚影,表现戏偶正在演出的一瞬间。 构图以**一幕一偶 + 大量留白**为核心。幕布是一片温暖的米黄或浅琥珀色光场,四周自然暗下,主体只占画面约25%–35%,可偏心或位于黄金分割位置,其余空间保持安静,像戏开场前的停顿。不出现观众、舞台全景或复杂道具。 配色从上方照片中提取 2–3 种最具情绪价值的颜色,转译为皮影彩染的半透明色,与暖灯光场统一;阴影使用深褐与暖墨色,而不是冷黑。 文字少量介入,可以竖排小字的方式写一个“戏名”或一句短句,像戏班手写的节目单,安静落在幕布一角,不限制语种。 整体呈现**皮影镂刻、半透明彩染、暖灯幕布、柔和投影与东方留白**共同构成的光影视觉。核心原则是:把照片里的主角请上幕布,让它用一个姿态演完一整出戏。避免完整写实、满场景堆砌、廉价国潮、塑料质感、3D渲染、卡通感和模板化海报。
点击高亮部分即可直接修改。
适用场景
Use when you have a photo of an animal, person, building, plant or object and want an elegant East-Asian exhibition-style poster pairing the photo with a shadow-puppet version of it.
怎样效果最好
- Upload your photo(s) together with the prompt; each photo becomes its own poster, they are not combined into one image.
- Works with any subject — the prompt treats animals, buildings or objects as a 'puppet character' with jointed parts and rods.
- Change the aspect ratio, the translucent dye colors or how much of the screen the puppet fills as single values; keep the lighting paragraph, which carries the look.
避坑提醒
- The model may change the top photo; restate that it must stay unaltered if the subject or composition shifts.
- Dense subjects can fill the whole screen; repeat the 25%–35% size and the empty-space rule if the puppet gets too big.
- The small vertical title text is generated and may be garbled Chinese characters.
相关模板
- MILO Penguin Apple Sled Comedy (Character Sheet → Seedance 2.5 Workflow)Use when you want a consistent original 3D animal character in a physics-driven slapstick short and need the reference-sheet step that locks identity.
- Manga-Panel Music Video Edit from Two Clips (Claude Opus 5.5)Use when you have two AI-generated (or your own) action clips and a music file and want an AMV / anime-opening style edit built by Claude with code, no editing software.
- Restaurant Kitchen Martial-Arts Fight Scene (Seedance 2.5, 10s Beat List)Use when you want a short, cinematic action beat with clear second-by-second choreography and a richly dressed real-world location.
- Character Goodnight Bedroom Visual (ChatGPT Image, Japanese Prompt)Use when you want a soft, personal bedtime illustration of your own original character that reflects their personality down to the pajamas and bedroom decor.
- Character Fashion Magazine Cover (ChatGPT Image, Japanese Prompt)Use when you have an original character and want a polished, magazine-style key visual that shows them in a fresh, modern outfit.
- Gunslinger vs Giant Slow-Mo Sandevistan Shot List (Seedance 2.5)Use when you want a stylised action beat with a slow-motion afterimage ('Sandevistan') effect, a rewind and a full-speed replay, all inside one Seedance generation.
来源与归属
- 示例案例作者
- DAAI